Re: [AMBER] MMPBSA.py.MPI and pytraj bugs

From: Hai Nguyen <nhai.qn.gmail.com>
Date: Sat, 2 Apr 2022 22:26:49 -0400

On Sat, Apr 2, 2022 at 8:46 PM David Poole <thepoole.ucdavis.edu> wrote:

> Cross-talk was indeed the issue. After some poking around I noticed that
> mpi4py was installed in my conda environment, but not the conda environment
> for amber (which is used by MMPBSA.py). To solve this issue, I only had to
> deactivate all conda environments and then use 'amber.conda' to install
> mpi4py into amber's conda environment. This seems to solve the problem I
> faced. It seems silly not to include mpi4py in amber's conda environment by
> default.



> i think the conda version is for serial install, not for MPI.

Hai

>
> I note that when using amber.python, and my conda environment (with mpi4py)
> activated, I could import mpi4py. But this was not the case with conda
> environments deactivated. I am not the exper to figure out the role of this
> cross talk, but it might be useful to note it.
>
> Anyway, thanks for the help, hope it continues to be useful to others.
>
> ~David Poole
>
> Can you double-check your PYTHONPATH, and report what it is pointing to?
> > I'm concerned that there may be some crosstalk between your "current
> > (conda)
> > environment" and what you call the "amber environment".
>
>
> > You can try by hand to execute the imports at the top of
> > $AMBERHOME/AmberTools/src/mmpbsa_py/MMPBSA.py. Basically, at line 53, it
> > is just doing "from mpi4py import MPI". Try this with amber.python under
> > circumstances where you know that the PYTHONPATH is the amber one. The
> > error message might be helpful.
>
>
> > ....regards...dac
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Apr 02 2022 - 19:30:02 PDT
Custom Search